ceph-ansible/infrastructure-playbooks/shrink-osd.yml

---
# This playbook shrinks Ceph OSDs.
# It can remove any number of OSD(s) from the cluster and ALL THEIR DATA
#
# Use it like this:
# ansible-playbook shrink-osd.yml -e osd_id=0,2,6
#     Prompts for confirmation to shrink, defaults to no and
#     doesn't shrink the cluster. yes shrinks the cluster.
#
# ansible-playbook -e ireallymeanit=yes|no shrink-osd.yml
#     Overrides the prompt using -e option. Can be used in
#     automation scripts to avoid interactive prompt.


- name: confirm whether user really meant to remove osd(s) from the cluster

  hosts:
    - localhost

  gather_facts: false
  become: true

  vars_prompt:
    - name: ireallymeanit
      prompt: Are you sure you want to shrink the cluster?
      default: 'no'
      private: no

  tasks:
  - include_vars: roles/ceph-common/defaults/main.yml
  - include_vars: group_vars/all.yml

  - name: exit playbook, if user did not mean to shrink cluster
    fail:
      msg: "Exiting shrink-osd playbook, no osd(s) was/were removed..
         To shrink the cluster, either say 'yes' on the prompt or
         or use `-e ireallymeanit=yes` on the command line when
         invoking the playbook"
    when: ireallymeanit != 'yes'

  - name: exit playbook, if no osd(s) was/were given
    fail:
      msg: "osd_ids must be declared
        Exiting shrink-osd playbook, no OSD()s was/were removed.
         On the command line when invoking the playbook, you can use
         -e osd_ids=0,1,2,3 argument."
    when: osd_ids is not defined

  - name: test if ceph command exist
    command: command -v ceph
    changed_when: false
    failed_when: false
    register: ceph_command

  - name: exit playbook, if ceph command does not exist
    debug:
      msg: "The ceph command is not available, please install it :("
    run_once: true
    when:
      - ceph_command.rc != 0

  - name: exit playbook, if cluster files do not exist
    stat:
      path: "{{ item }}"
    register: ceph_conf_key
    with_items:
      - /etc/ceph/{{ cluster }}.conf
      - /etc/ceph/{{ cluster }}.client.admin.keyring
    failed_when: false

  - fail:
      msg: "Ceph's configuration file is not present in /etc/ceph"
    with_items: "{{ ceph_conf_key.results }}"
    when:
      -  item.stat.exists == false

  - name: exit playbook, if can not connect to the cluster
    command: timeout 5 ceph --cluster {{ cluster }} health
    register: ceph_health
    until: ceph_health.stdout.find("HEALTH") > -1
    retries: 5
    delay: 2

# NOTE (leseb): just in case, the complex filters mechanism below does not work anymore.
# This will be a quick and easy fix but will require using the shell module.
#  - name: find the host where the osd(s) is/are running on
#    shell: |
#      ceph --cluster {{ cluster }} osd find {{ item }} | grep -Po '(?<="ip": ")[^:]*'
#    with_items: "{{osd_ids.split(',')}}"
#    register: osd_hosts
#
  - name: find the host where the osd(s) is/are running on
    command: ceph --cluster {{ cluster }} osd find {{ item }}
    with_items: "{{ osd_ids.split(',') }}"
    register: osd_hosts

  - set_fact: ip_item="{{(item.stdout | from_json).ip}}"
    with_items: "{{ osd_hosts.results }}"
    register: ip_result

  - set_fact: ips="{{ ip_result.results | map(attribute='ansible_facts.ip_item') | list }}"

  - set_fact: real_ips="{{ ips | regex_replace(':[0-9][0-9][0-9][0-9]\/[0-9]+', '') }}"

  - name: check if ceph admin key exists on the osd nodes
    stat:
      path: "/etc/ceph/{{ cluster }}.client.admin.keyring"
    register: ceph_admin_key
    with_items: "{{ real_ips }}"
    delegate_to: "{{ item }}"
    failed_when: false

  - fail:
      msg: "The Ceph admin key is not present on the OSD node, please add it and remove it after the playbook is done."
    with_items: "{{ ceph_admin_key.results }}"
    when:
      -  item.stat.exists == false

  - name: deactivating osd(s)
    command: ceph-disk deactivate --cluster {{ cluster }} --deactivate-by-id {{ item.0 }} --mark-out
    register: deactivate 
    ignore_errors: yes
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ real_ips }}"
    delegate_to: "{{ item.1 }}"

  - name: set osd(s) out when ceph-disk deactivating fail
    command: ceph --cluster {{ cluster }} osd out osd.{{ item.0 }}
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ deactivate.results }}"
    when: 
      - item.1.stderr|length > 0

  - name: destroying osd(s)
    command: ceph-disk destroy --cluster {{ cluster }} --destroy-by-id {{ item.0 }} --zap
    register: destroy 
    ignore_errors: yes
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ real_ips }}"
    delegate_to: "{{ item.1 }}"

  - name: remove osd(s) from crush_map when ceph-disk destroy fail
    command: ceph --cluster {{ cluster }} osd crush remove osd.{{ item.0 }}
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ destroy.results }}"
    when: 
      - item.1.stderr|length > 0

  - name: delete osd(s) auth key when ceph-disk destroy fail 
    command: ceph --cluster {{ cluster }} auth del osd.{{ item.0 }}
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ destroy.results }}"
    when: 
      - item.1.stderr|length > 0

  - name: deallocate osd(s) id when ceph-disk destroy fail 
    command: ceph --cluster {{ cluster }} osd rm {{ item.0 }}
    with_together:
      - "{{ osd_ids.split(',') }}"
      - "{{ destroy.results }}"
    when: 
      - item.1.stderr|length > 0